Augos Factory / Learn / Kaizen
Learn · kaizen and continuous improvement
Kaizen works best when every stop is already on record.
In short
Kaizen is continuous improvement through small, frequent changes made by the people who do the work. Each cycle picks one problem, measures it, changes the method and checks the result (plan, do, check, act). It works best when the team starts from a complete record of stops, not from memory.
The basics
What is kaizen?
Kaizen is Japanese for "change for the better". The word joins kai (change) and zen (good). In a plant it means improvement as a habit: many small changes, tested by the people on the line, rather than occasional large projects.
The idea took shape in Japanese factories after the Second World War. American job-methods training and the teaching of W. Edwards Deming both fed into it. Toyota built it into its production system, where every operator is expected to spot problems. Masaaki Imai's book, Kaizen, later carried the word to the West.
It runs in two modes. Daily kaizen is the small fix a line team makes and keeps. A kaizen event is a focused effort of a few days on one problem: a slow changeover, a recurring jam, an overrunning clean.
Both follow the same cycle: plan what to change, do it, check whether it worked, and act to make it the new standard. Kaizen sits inside lean: lean is the wider system, kaizen the daily habit.
Principles
The five principles of kaizen
Lists vary between authors, but most guides share these five.
| Principle | What it means | On the line |
|---|---|---|
| Know your customer | Improve what the customer values: quality, delivery and cost. | Start with losses that cost output or quality, not with tidying for its own sake. |
| Let it flow | Remove waste so work moves without waiting. | Every wait for material, a technician or a sign-off is a break in flow. |
| Go to gemba | Gemba is Japanese for the real place: the shop floor where the work is done. Decide there, with the facts. | Watch the changeover itself, not the report about it. |
| Involve the people who do the work | Operators and artisans know the task best. They suggest and test the changes. | The crew with the fastest changeover teaches its method to the others. |
| Be transparent | Make performance visible, so problems surface early. | Show the line's status and today's losses where the team can see them. |
What kaizen targets
Three kinds of loss kaizen removes
The Toyota Production System names three kinds of loss. Muda is waste: work the customer would not pay for. Mura is unevenness. Muri is overburden: people or machines pushed too hard.
Muda is often split into seven wastes: overproduction, waiting, transport, over-processing, inventory, motion and defects.
Waiting is the waste a stop record sees most directly. On one anonymised beverage line, waits for raw material took 97.8 h of 720.0 paid hours in one month: 13.58%.
Mura shows up as spread: the same clean taking 103 minutes one day and 204 the next.
Two modes
Daily kaizen and kaizen events compared
| Question | Daily kaizen | Kaizen event |
|---|---|---|
| Who takes part? | The operators on the line and their supervisor | A small cross-functional team, chosen for the problem |
| How long? | Minutes a day, every day | Often three to five days, then weeks of checks |
| Typical target | Interruptions (short stops), small jams, a messy station | A slow changeover, a recurring fault, an overrunning clean |
| What it leaves behind | A small change, kept | A new standard method, written and trained |
A plant needs both: daily kaizen for the many small stops, an event for the few long ones.
The record shows which is which. In one week on one line, 3,175 of 3,334 stops were interruptions, taking 14.8 h. The other 159 stops took 23.8 h.
Interruptions add up. Across 50 South African lines they took 10.1% of paid time, pooled, and 22 of the 50 lines lost 10% or more.
One week on one line
MeasuredAugos platform, Downtime Report · one anonymised line · one week in 2026 · site, products and people removed. Longer stops are all stops less interruptions: 3,334 less 3,175 is 159; 38.6 h less 14.8 h is 23.8 h.
The cycle
How the PDCA cycle runs on a line
PDCA (plan, do, check, act) is the engine of kaizen, taught in Japan by W. Edwards Deming. Each turn of the cycle is one small experiment.
The check step is the easiest to skip, and the only one that shows the change held.
Plan
Choose one problem and measure it as it is. Write down the change and what success looks like.
Do
Try the change small: one crew, one product or one shift.
Check
Measure the same thing the same way. Compare the result with the baseline, not with memory.
Act
If it worked, write it into standard work and train every shift. If not, keep the lesson and plan the next try.
Why events stall
Six ways a kaizen event loses its way
The loudest problem, not the biggest
Without a ranked list of losses, the team works on whatever was argued about in the last meeting.
No baseline
If nobody measured the task before, nobody can show the change worked.
The gain fades
Three months later the old method is back, and there is no record to show when it slipped.
Counting instead of timing
A ranking by count puts the many interruptions first and hides the few long ones. Rank by hours.
Fixing before measuring
The team agrees a fix in the first hour. Nobody checks whether it touches the loss that matters.
Too wide a scope
"Improve the packing hall" is too wide. "Cut CIP time on line 2" is a kaizen event: one task, one line, one measure.
A kaizen event, with the record
Five steps, each with a number
The fastest version of a task is usually already in your data. Kaizen with a record starts there.
Pick the biggest loss by hours
Rank stop reasons by hours over the last month, not by how often they happen.
Set the baseline
The typical time: the median of that task's events over a period before the event. Write it down before anyone changes anything.
Study the fast runs
What was different on the good days? Which team, which product, which shift, which tools? Then go to the line and watch the next one.
Write it into standard work
Make the fast version the method, with the people who do it.
Check the next weeks
Run the same report on the weeks after the event, at least ten events. If the median has moved towards the recommended time, the change held.
Run kaizen on the record
The same report, before and after
A kaizen event needs a baseline before anything changes, and the same measure afterwards. The Planned Stoppage Analysis gives both: run it on the weeks before the event, then on the weeks after.
The median comes from the period you pick. The recommended time comes from the calendar year to date: the median less the median absolute deviation, outliers set aside. The gap between them is the variation the event works on.
When the line's best beats the allocated time, planning can tighten the allocated time. The platform shows the gap; the plant sets the new allocated time.
Planned Stoppage Analysis · one changeover · before and after
IllustrativeA median of 40 min against a recommended time of 22. The fast version exists on this line. It just isn’t repeated.
Illustrative stop, not a customer. Recoverable hours are each changeover's time above the recommended time, summed; here approximated as (median − recommended) × changeovers ÷ 60.
In the toolbox
From the record
The five steps on a real cleaning cycle
Cleaning in place (CIP) suits kaizen: it repeats often, it has an allocated time, and crews run it differently. Here is one month on one anonymised beverage line.
CIP on one beverage line, one month
Measured| Measure | Value |
|---|---|
| CIP cycles | 24 |
| Allocated per cycle | 170 min |
| Cycles over the allocated time | 13 |
| Time over the allocated time | 7.6 h |
| Fastest, outliers set aside | 103 min |
| Slowest, outliers set aside | 204 min |
| Recommended time (median less MAD) | 2 h 14 min |
| Time above the recommended time | 16.9 h |
Augos platform, Planned Stoppage Analysis · one anonymised beverage line · one month in 2026. Three cycles under 59 min and two over 271 min were set aside as outliers. Site, products and people removed.
Pick the loss
Cleaning is a planned stop, so it is easy to treat as fixed. The record says otherwise: 13 of 24 cycles ran over the 170-minute allocated time.
Set the baseline
7.6 h is 456 minutes. Over 13 cycles, that is about 35 minutes of overrun each.
Study the fast cycles
The line's recommended time is 2 h 14 min, 36 minutes inside the allocated time. Was the line drained on time? Were the chemicals at temperature? Who was on shift?
Write the standard
Put the fast cycle's sequence into the method, with the crew. Where the line's best beats the allocated time, planning can tighten the allocated time.
Check the next weeks
Run the same report on the weeks after the event, at least ten cycles. At 24 cycles a month, that takes about two weeks. If the median has moved towards the recommended time, the change held.
These figures are a starting point, not a result: 16.9 h above the line's recommended time in one month.
The toolkit
Kaizen tools and when to use them
| Tool | What it does | Use it when |
|---|---|---|
| Pareto by hours | Ranks losses by total hours, not by count | Choosing what to work on first |
| Gemba walk | Leaders watch the work at the line and ask questions | Before and after a change |
| 5S | Sort, set in order, shine, standardise, sustain | Searching for tools or parts slows the task |
| Standard work | The current best method, written with the people who do it | Locking in a gain for every shift |
| 5 whys | Asks why until the answer is something the team can change | One clear problem with a short causal chain |
| Fishbone (Ishikawa) | Sorts possible causes under method, machine, material, people, measurement and environment | Several possible causes need testing |
| SMED | Moves changeover work outside the stop, then streamlines what remains | Changeovers are the biggest planned loss |
| Visual management | Shows status and losses where the team can see them | Daily kaizen needs fast feedback |
Who does what
Roles in a kaizen event
Keep the team small and chosen for the problem.
- Sponsor. A manager who chooses the problem, clears obstacles and signs off the new standard.
- Facilitator. Keeps the event on the method: baseline first, change second, check third.
- Operators and artisans. They do the task, know where the time goes and test each change.
- Supervisor. Owns the standard after the event and checks it on every shift.
- Data lead. Reads the record, sets the baseline and runs the check.
- Maintenance, quality or planning. Join when the loss sits in their area.
Sustaining the gain
How to make a kaizen gain stick
Most kaizen gains fade for one reason: nobody measures the task after the event. Old habits return quietly, one shift at a time.
- Write the standard with the crew, at the station.
- Train every shift, not only the one in the event.
- Keep the same measure running: the same report, run every period.
- Review it at the shift meeting, with the line's status in view.
- Let the target move as the typical time falls.
- Start the next cycle. Kaizen is a habit, not a project.
If the typical time drifts back up, the record shows when, which shift and which product.
Why the record matters
What a complete stop record changes at each step
| Step | From memory | From a complete record |
|---|---|---|
| Pick the problem | The loss argued about last | The biggest loss by hours, ranked |
| Set the baseline | A guess, or a stopwatch on one good day | The median of every event in the period, timed by a sensor |
| Find the fast version | Opinions about who is quickest | The fast runs, with team, product and shift |
| Write the standard | A target from a textbook | A target the line has already reached |
| Check the result | One good week after the event | The same measure on the weeks after, at least ten events, and every period after |
Questions
Kaizen, answered
What does kaizen mean?
Kaizen is Japanese for "change for the better". In business it means continuous improvement through small, frequent changes made by everyone.
What are the 5 principles of kaizen?
Know your customer, let it flow, go to gemba, involve the people who do the work, and be transparent. Lists vary between authors.
What is the kaizen method?
Pick one problem, measure it, test a change and check the result. If it works, it becomes the new standard: plan, do, check, act.
What are the 10 rules of kaizen?
A widely shared list runs like this. Discard fixed ideas, and think about how it can be done, not why it cannot. Make no excuses and do not wait for perfection. Correct mistakes at once. Spend ideas before money. Treat problems as chances to learn. Ask why five times. Seek ideas from many people, not one. Improvement never ends.
What is a good kaizen example?
A task that repeats against an allocated time, such as a changeover or a clean. On one beverage line, the same CIP clean took 103 minutes at its fastest and 204 at its slowest. Kaizen makes the fast version the method.
What is the difference between kaizen and lean?
Lean is the wider system of removing waste from a process. Kaizen is the habit of continuous, small improvement that lean relies on.
How long is a kaizen event?
Often three to five days on one problem, with the people who do the work, followed by checks over the following weeks.
Start your next kaizen event with the facts.
Thirty minutes with a manufacturing specialist: every stop on a line ranked by hours, so the team works on the biggest loss first.
- Every stop timed, short ones included
- A recommended time for each task
- Prices in rand, ex VAT
A manufacturing specialist replies within two business hours to set a 30-minute call.